Elastic search Synonym completely eliminated by analyzer

Question

I am using synonym file to create synonyms in elasticsearch, My requirement is to show photo frames of different sizes.

For example-

6x9, 6 x 9 => 6x9

But when I close and re-open the index, I am getting following error.

{
"error": {
    "root_cause": [
        {
            "type": "illegal_argument_exception",
            "reason": "failed to build synonyms"
        }
    ],
        "type": "illegal_argument_exception",
            "reason": "failed to build synonyms",
                "caused_by": {
        "type": "parse_exception",
            "reason": "Invalid synonym rule at line 107",
                "caused_by": {
            "type": "illegal_argument_exception",
                "reason": "term: 6 x 9 was completely eliminated by analyzer"
        }
    }
},
"status": 400
}

It works fine for

8x10, 8 x 10 => 8x10

Which means it is only working if it has minimum 2 digits after x ie. 10 in 8 x 10 . Regarding 6x9 it is working fine. The only issue is with 6 x 9 as it has spaces and the last digit is single. But it is working fine if I change it to 6 x 09.

Here are the Settings -

"analysis": {
"filter": {
    "synonym_filter": {
        "type": "synonym",
            "synonyms_path": "analysis/synonyms.txt"
    },
    "suggestions_shingle": {
        "max_shingle_size": "4",
            "min_shingle_size": "2",
                "type": "shingle"
    },
    "english_stemmer_filter": {
        "name": "minimal_english",
            "type": "stemmer"
    },
    "edgeNGram_filter": {
        "min_gram": "2",
            "side": "front",
                "type": "edgeNGram",
                    "max_gram": "20"
    }
},
"analyzer": {
    "whitespace_punc_analyzer": {
        "filter": [
            "lowercase",
            "asciifolding",
            "word_delimiter"
        ],
            "type": "custom",
                "tokenizer": "whitespace"
    },
    "edge_nGram_analyzer": {
        "filter": [
            "lowercase",
            "asciifolding",
            "synonym_filter"
        ],
            "type": "custom",
                "tokenizer": "edge_ngram_tokenizer"
    },
    "path_analyzer_lc": {
        "filter": [
            "lowercase"
        ],
            "tokenizer": "path_tokenizer"
    },
    "stemmer_synonym_analyzer": {
        "filter": [
            "synonym_filter",
            "lowercase",
            "english_stemmer_filter"
        ],
            "tokenizer": "standard"
    },
    "whitespace_analyzer": {
        "filter": [
            "lowercase",
            "asciifolding"
        ],
            "type": "custom",
                "tokenizer": "whitespace"
    },
    "synonym_analyzer": {
        "filter": [
            "synonym_filter",
            "lowercase",
            "edgeNGram_filter"
        ],
            "tokenizer": "standard"
    },
    "edge_nGram_shingle_analyzer": {
        "filter": [
            "lowercase",
            "asciifolding",
            "synonym_filter",
            "suggestions_shingle"
        ],
            "type": "custom",
                "tokenizer": "edge_ngram_tokenizer"
    },
    "path_analyzer": {
        "tokenizer": "path_tokenizer"
    }
},
"tokenizer": {
    "edge_ngram_tokenizer": {
        "token_chars": [
            "letter",
            "digit"
        ],
            "min_gram": "2",
                "type": "edgeNGram",
                    "max_gram": "6"
    },
    "path_tokenizer": {
        "ignore_case": "true",
            "type": "path_hierarchy",
                "delimiter": ">"
    }
}}

Thanks in advance!

Answer 1

It's because of the edge_ngram_tokenizer tokenizer that has min_gram set to 2, and hence, it cannot produce any tokens for single character input.

POST _analyze
{
  "text": "6 x 9",
  "tokenizer": {
    "token_chars": [
      "letter",
      "digit"
    ],
    "min_gram": "2",
    "type": "edgeNGram",
    "max_gram": "6"
  }
}

=> tokens: []

For 8 x 10, only the token 10 is produced, which is probably not what you want either:

POST _analyze
{
  "text": "8 x 10",
  "tokenizer": {
    "token_chars": [
      "letter",
      "digit"
    ],
    "min_gram": "2",
    "type": "edgeNGram",
    "max_gram": "6"
  }
}

=> tokens: [10]

So the reason you're getting this error message is because the tokenizer doesn't produce any tokens, and then token filters have nothing to chew on.